SCIENTIFIC EVIDENCE IN HAZARD DETERMINATION
Scientific information about the biological effects of a substance or agent generally consists of an interrelated network of theoretical, experimental, circumstantial, and epidemiological evidence that must be assessed as a whole. In addition to the direct evidence about biological effects, it is sometimes also necessary to consider other evidence about the properties of the substance or agent that may contribute to its potential hazard (e.g., solubility, volatility, flammability, electrical conductivity under different conditions, and, in the case of chemical pollutants, the mechanism by which the substance is removed from the biosphere). For a given potential hazard, certain types of evidence may be absent or equivocal, and in some circumstances more weight may be given to one type of evidence than to the others. Nonetheless, all available evidence should be considered.
Theoretical evidence is concerned with mechanisms of interaction between the substance or agent and the biological tissues or system under consideration, and with the chemical or physical properties of the substance or agent that might cause a biological effect. In the hearings, the theoretical evidence is sometimes referred to as biophysical evidence or biophysical theory. It consists of testimony regarding the ability of the electric fields under the power lines to cause tissue heating or molecular polarization or deformity within tissues. The evidence is generally valid for what it states: that electric fields cannot produce sufficient heating or molecular polarization to cause significant biological effects. Note, however, that other as yet unknown biophysical mechanisms could exist by which a biological effect could be produced by electric fields, and these mechanisms are not covered by existing theory. Hence, the absence of theoretical explanation does not necessarily imply absence of effect.
Experimental evidence results from direct exposure and observation of experimental animals arid sometimes of man. Experimental evidence is usually considered stronger than theoretical evidence, but it is subject to various limitations. One such limitation concerns the adequacy of experimental design and the validity of conclusions. Another limitation is that studies of biological responses to chemical and physical agents center on biological effect rather than hazard. In fact, the experimental evidence for an actual hazard may be inconclusive, even though the experiments have been conducted properly. In the hearing, the experimental results were largely obtained from studies unrelated to hazard determination. For instance, Marino states that his studies were intended to find ways to promote healing of bone fractures. By contrast, the original studies on hazards of ionizing radiation were designed to reveal potential hazard. The results of the studies showed that radiation levels above 0.1 R/day1 reduces the life span of the animals, an effect that can be presumed to imply hazard.
Circumstantial evidence relies on the known effects of similar or related substances or agents, or similar effects in other animal species. Thus, if a chemical compound is known to cause cancer, another compound with a related chemical structure might also be suspected of causing cancer. In the broad sense, all experimental results in animals present circumstantial evidence for effects in humans. Circumstantial evidence about hazards to humans does not prove actual hazard, but regulatory agencies acting in the public interest frequently accept results from animal studies in deciding about human hazard.
Epidemiological evidence derives from monitoring health indices in a human population that is exposed to a substance or agent. Epidemiological studies are usually health-oriented, and provide direct evidence about potential human hazard under the conditions of exposure. These studies, however, are subject to a number of problems that limit the validity of their conclusions. They are usually retrospective; appropriate controls are difficult to obtain, and the results are often presented in the form of a correlation ('A is associated with B"), rather than a direct cause-and-effect relationship. For these reasons, epidemiological results are inherently weak, and should be supported by experimental evidence.
Other aspects of scientific evidence as well must be considered in evaluating hazards to humans:
Confirmation versus duplication. An experiment performed a sufficient number of independent times or on a sufficient number of animals with consistent results establishes beyond reasonable doubt that a true effect has occurred. Beyond that point it is not usually considered necessary to repeat the experiment unless one suspects an artifact or a flaw in the experimental design. Independent confirmation in several laboratories is particularly important in developing a scientific consensus. Ultimate acceptance of the experimental results depends on how they fit into the general body of scientific knowledge. When an experiment is repeated, the experiment is often redesigned to yield additional information. For instance, the original experiments on the hazard of chronic exposure to ionizing radiation were repeated, but the new experiments were redesigned to yield data from which a dose-rate/ effect formula could be derived.
In the hearings, the validity of the experimental results presented by Marino were challenged in cross examination. Considering the points raised and the general nature of the results, the experiments should probably be repeated, with design modifications that would give more information about hazard.
The Extrapolation Problem. Testimony at the hearings was related to the question of whether or not a hazard exists, but this is not the question asked by a regulatory body involved in setting the permissible conditions of exposure of humans to avoid or minimize human hazard. For power lines, this might involve questions of width of right-of-way, height of power lines, or shielding. The extrapolation problem involves transferring data on biological effects from the circumstances in which they were acquired to conditions under which man might be exposed, and then judging what might happen to man. The problem includes transferring data from animals to man, predicting effects at a low concentration or intensity from results obtained at a high concentration or intensity, and predicting the effects for a particular kind of agent from results obtained in a physically or chemically related agent.
Predicting how biological effects in man depend on concentration, amount, or duration of exposure to a substance is difficult, particularly when the data are derived from other animal species. For effects that depend on the interaction of a substance with normal biochemical and physiological processes, equivalent effects in different species are produced when the amount or concentration of the substance is proportional to the body surface area of the animals. For other effects, the amount may be proportional to the body weight of the animals or to the rate of some physiological process such as respiratory exchange or urine excretion. When the biological effects of a substance are believed to be completely reversible, the usual practice is to estimate the level for human effects from animal data, reduce this level by some safety factor, and test the reduced level in human volunteers for absence of effect. The size of the safety factor varies: for industrial gases, atmospheric contaminants, and prescription drugs, the safety factor ranges from 2 to 20; for food additives, it is 100.
For exposure to microwaves the safety factor was originally chosen to be 10, but more recent work indicates that the allowed level implies a safety factor of between 4 and 6. For ionizing radiation, the safety factor excludes medical irradiation for diagnosis or therapy, which adds a variable amount to the radiation burden of each individual.
The most serious questions about extrapolation of biological effects to low dose or concentration involve irreversible effects (e.g., carcinogenesis). For carcinogenesis, the relationship between the amount or concentration of a cancer-producing substance and the degree of effect produced at low exposure levels is unknown. To deal with the data on cancer induction in rats and mice, several statistical procedures have been proposed. Using conservative assumptions, these procedures attempt to estimate the concentration or amount of the substance that will produce no more than some arbitrarily low level of cancer cases in the human population-usually in I out of 1,000,000 people. The procedures extrapolate from cancer rates of 10 to 40% observed in mice down to a 0.0001% rate predicted in humans in the absence of any parametric model. This arbitrarily low rate of occurrence is adopted by regulatory boards for want of any better way to reach a decision.
1 R/day = Roentgen per day.
Marino Home Page | Research Interests